View : 198 Download: 0

Full metadata record

DC Field Value Language
dc.contributor.advisor김수영-
dc.contributor.author전민정-
dc.creator전민정-
dc.date.accessioned2018-09-05T08:24:19Z-
dc.date.available2018-09-05T08:24:19Z-
dc.date.issued2018-
dc.identifier.otherOAK-000000151200-
dc.identifier.urihttp://dcollection.ewha.ac.kr/common/orgView/000000151200en_US
dc.identifier.urihttp://dspace.ewha.ac.kr/handle/2015.oak/245788-
dc.description.abstract다지표 잠재성장모형(second-order latent growth model)은 측정변수들의 요인구조를 모형에 포함하여 잠재요인의 종단적 변화를 설명하는 발전된 형태의 잠재성장모형이다. 다지표 잠재성장모형은 단일지표 모형에 비해 여러 장점을 갖는 유용한 모형이지만, 자료에 부분측정동일성이 존재하는 경우에는 잠재요인의 척도를 설정하는 것에 의해 모수추정이 영향을 받는 문제가 있다. 이 때 요인척도설정 방식(참조변수, 효과코딩)에 따라 모수추정에 발생하는 편향의 정도가 달라질 수 있어 이를 최소화하는 방식에 대한 탐색이 중요하다. 이에 본 연구는 시뮬레이션을 통해 다양한 부분측정동일성 상황에서 두 요인척도설정 방식에 따라 다지표 잠재성장모형의 모수추정치들이 받는 영향을 확인하였다. 시뮬레이션 결과, 요인의 척도설정을 위해 제약되는 문항(scaling item)이 측정동일성을 만족하는 경우에는 참조변수 방식이 유리하나, 제약되는 문항이 측정동일성을 만족하지 못하는 경우에는 효과코딩 방식이 모수추정에 발생하는 편향을 최소화하는 것을 확인하였다. 본 연구의 결과를 바탕으로, 실제 종단자료를 다지표 잠재성장모형으로 분석하고자 할 때 요인척도설정 방식을 어떻게 적용할 수 있는지에 대해 논의하였다.;Latent growth model has been widely applied for longitudinal data to identify the change of a construct of interest over time. Recently, second-order latent growth model has received much attention for its advantages in modeling longitudinal changes over the traditionally used first-order latent growth model. Although the second-order latent growth model has been especially highlighted for the capability of modeling various levels of longitudinal measurement invariance, the growth parameter estimates of the model could possibly be affected by the scaling of latent factor when partial measurement invariance is implied in data. However, the extent to which the parameter estimates are affected could differ from scaling methods: marker variable method and effects coding method. Therefore, investigating the relative performance of the two scaling methods in the context of second-order latent growth model is important for better understanding of the model and its applications. The purpose of the present study is to examine and compare the performance of second-order latent growth model with the two scaling methods under various partial measurement invariance settings. To achieve the goal, an extensive Monte Carlo simulation study was conducted, and the performance of the second-order latent growth model under the two scaling approaches were evaluated in terms of the parameter estimates accuracy. The results showed that, with the ideal conditions in which the item restricted for the scaling is invariant over time, the marker variable method provided more accurate parameter estimates than the effects coding method. However, with the non-invariance in the restricted item, the effects coding method displayed more accurate parameter estimates across all partial measurement invariance settings. A discussion of the results and limitations of the study are provided as well as general recommendations.-
dc.description.tableofcontentsI. Introduction 1 II. Theoretical Framework 5 A. Second-order latent growth model 5 B. Full and Partial Measurement Invariance in the Longitudinal Context 7 C. Scaling Methods and Scaling-related Misspecification Issue in SLGM 9 III. Method 14 A. Simulated Data Conditions 14 B. Population Models for Data Generation 16 C. Evaluation Measures for the Results 19 IV. Results 20 A. Model Non-convergence Problem 20 B. Overall Comparisons of the Accuracy of the Estimation 21 C. Comparisons of the Stability of the Estimation 27 D. Influence of the Size of Lambda 29 V. Discussion and Conclusion 32 Reference 35 Abstract(in Korean) 41-
dc.formatapplication/pdf-
dc.format.extent1081875 bytes-
dc.languageeng-
dc.publisher이화여자대학교 대학원-
dc.subject.ddc100-
dc.titlePerformance of Second-order Latent Growth Model under Partial Longitudinal Measurement Invariance-
dc.typeMaster's Thesis-
dc.title.subtitleA Comparison of Two Scaling Approaches-
dc.title.translated부분측정동일성 상황에서 다지표 잠재성장모형의 모수추정 정확성 연구 : 두 요인척도설정 방식의 영향력 비교-
dc.creator.othernameMin Jeong Jeon-
dc.format.pageiv, 41 p.-
dc.contributor.examiner김성호-
dc.contributor.examiner양수진-
dc.contributor.examiner김수영-
dc.identifier.thesisdegreeMaster-
dc.identifier.major대학원 심리학과-
dc.date.awarded2018. 8-
Appears in Collections:
일반대학원 > 심리학과 > Theses_Master
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE